public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30)
@ 2023-08-10 13:33 hubicka at gcc dot gnu.org
2023-08-10 13:45 ` [Bug middle-end/110972] " hubicka at gcc dot gnu.org
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: hubicka at gcc dot gnu.org @ 2023-08-10 13:33 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
Bug ID: 110972
Summary: 13% fatigue regression between g:bb3ceeb6520c13fc
(2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08
13:30)
Product: gcc
Version: 13.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: hubicka at gcc dot gnu.org
Target Milestone: ---
With -Ofast -march native.
zen2:
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=171.767.0
zen3:
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=283.767.0
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug middle-end/110972] 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30)
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
@ 2023-08-10 13:45 ` hubicka at gcc dot gnu.org
2023-08-11 7:05 ` rguenth at gcc dot gnu.org
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: hubicka at gcc dot gnu.org @ 2023-08-10 13:45 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
--- Comment #1 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
The following patches in the range looks like they may cause the difference
commit d9f3ea61fe36e2de3354b90b65ff8245099114c9
Author: Richard Biener <rguenther@suse.de>
Date: Mon Aug 7 14:44:20 2023 +0200
tree-optimization/49955 - BB reduction with odd number of lanes
commit 02a015fa320229a057ef721eaf663f3eb22a8ace
Author: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
Date: Mon Aug 7 17:38:12 2023 +0800
VECT: Support CALL vectorization for COND_LEN_*
commit 31ec413098bd334115aff73fc755e49afd3ac371
Author: Richard Biener <rguenther@suse.de>
Date: Tue Aug 8 12:46:42 2023 +0200
tree-optimization/110924 - fix vop liveness for noreturn const CFG parts
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug middle-end/110972] 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30)
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
2023-08-10 13:45 ` [Bug middle-end/110972] " hubicka at gcc dot gnu.org
@ 2023-08-11 7:05 ` rguenth at gcc dot gnu.org
2023-08-11 8:24 ` rguenth at gcc dot gnu.org
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: rguenth at gcc dot gnu.org @ 2023-08-11 7:05 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2023-08-11
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot gnu.org
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
I'm quite sure it's the reduction vectorization.
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug middle-end/110972] 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30)
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
2023-08-10 13:45 ` [Bug middle-end/110972] " hubicka at gcc dot gnu.org
2023-08-11 7:05 ` rguenth at gcc dot gnu.org
@ 2023-08-11 8:24 ` rguenth at gcc dot gnu.org
2023-08-11 9:31 ` rguenth at gcc dot gnu.org
2024-04-04 7:40 ` [Bug middle-end/110972] 13% fatigue regression with r14-3078-gd9f3ea61fe36e2 rguenth at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: rguenth at gcc dot gnu.org @ 2023-08-11 8:24 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
I can confirm that on Zen2 r14-3078-gd9f3ea61fe36e2 regresses fatigue by 12%.
I'm going to take a look if there's sth obvious.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug middle-end/110972] 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30)
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
` (2 preceding siblings ...)
2023-08-11 8:24 ` rguenth at gcc dot gnu.org
@ 2023-08-11 9:31 ` rguenth at gcc dot gnu.org
2024-04-04 7:40 ` [Bug middle-end/110972] 13% fatigue regression with r14-3078-gd9f3ea61fe36e2 rguenth at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: rguenth at gcc dot gnu.org @ 2023-08-11 9:31 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
-fatigue2.f90:1077:93: optimized: basic block part vectorized using 32 byte
vectors
+fatigue2.f90:1002:56: optimized: basic block part vectorized using 32 byte
vectors
+fatigue2.f90:1058:93: optimized: basic block part vectorized using 32 byte
vectors
OTOH that's a bit imprecise, the - is bogus I think, but the + are real,
:1058 is also loop vectorized. The body is very large, perf shows
Samples: 279K of event 'cycles', Event count (approx.): 316851940248
Overhead Samples Command Shared Object Symbol
30.84% 86412 fatigue2 fatigue2 [.] MAIN__
24.26% 67530 fatigue fatigue [.] MAIN__
20.05% 56167 fatigue2 fatigue2 [.]
__perdida_m_MOD_generalized_hookes_law.constprop.0.isra.0
18.73% 52127 fatigue fatigue [.]
__perdida_m_MOD_generalized_hookes_law.constprop.0.isra.0
1.96% 5459 fatigue fatigue [.]
__perdida_m_MOD_generalized_hookes_law.constprop.1.isra.0
1.84% 5140 fatigue2 fatigue2 [.]
__perdida_m_MOD_generalized_hookes_law.constprop.1.isra.0
0.83% 2308 fatigue libc-2.31.so [.]
__memset_avx2_unaligned_erms
0.60% 1693 fatigue2 libc-2.31.so [.]
__memset_avx2_unaligned_erms
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug middle-end/110972] 13% fatigue regression with r14-3078-gd9f3ea61fe36e2
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
` (3 preceding siblings ...)
2023-08-11 9:31 ` rguenth at gcc dot gnu.org
@ 2024-04-04 7:40 ` rguenth at gcc dot gnu.org
4 siblings, 0 replies; 6+ messages in thread
From: rguenth at gcc dot gnu.org @ 2024-04-04 7:40 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110972
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|rguenth at gcc dot gnu.org |unassigned at gcc dot gnu.org
Status|ASSIGNED |NEW
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2024-04-04 7:40 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-08-10 13:33 [Bug middle-end/110972] New: 13% fatigue regression between g:bb3ceeb6520c13fc (2023-08-07 21:09) and g:d9dc70cc65becca9 (2023-08-08 13:30) hubicka at gcc dot gnu.org
2023-08-10 13:45 ` [Bug middle-end/110972] " hubicka at gcc dot gnu.org
2023-08-11 7:05 ` rguenth at gcc dot gnu.org
2023-08-11 8:24 ` rguenth at gcc dot gnu.org
2023-08-11 9:31 ` rguenth at gcc dot gnu.org
2024-04-04 7:40 ` [Bug middle-end/110972] 13% fatigue regression with r14-3078-gd9f3ea61fe36e2 rguenth at gcc dot gnu.org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).